2 * modified for EXT2FS support in Lites 1.1
4 * Aug 1995, Godmar Back (gback@cs.utah.edu)
5 * University of Utah, Department of Computer Science
8 * Copyright (c) 1989, 1991, 1993, 1994
9 * The Regents of the University of California. All rights reserved.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/namei.h>
44 #include <sys/kernel.h>
45 #include <sys/vnode.h>
46 #include <sys/mount.h>
50 #include <sys/fcntl.h>
51 #include <sys/malloc.h>
53 #include <sys/mutex.h>
55 #include <geom/geom.h>
56 #include <geom/geom_vfs.h>
58 #include <fs/ext2fs/ext2_mount.h>
59 #include <fs/ext2fs/inode.h>
61 #include <fs/ext2fs/fs.h>
62 #include <fs/ext2fs/ext2_extern.h>
63 #include <fs/ext2fs/ext2fs.h>
65 static int ext2_flushfiles(struct mount *mp, int flags, struct thread *td);
66 static int ext2_mountfs(struct vnode *, struct mount *);
67 static int ext2_reload(struct mount *mp, struct thread *td);
68 static int ext2_sbupdate(struct ext2mount *, int);
69 static int ext2_cgupdate(struct ext2mount *, int);
70 static vfs_unmount_t ext2_unmount;
71 static vfs_root_t ext2_root;
72 static vfs_statfs_t ext2_statfs;
73 static vfs_sync_t ext2_sync;
74 static vfs_vget_t ext2_vget;
75 static vfs_fhtovp_t ext2_fhtovp;
76 static vfs_mount_t ext2_mount;
78 MALLOC_DEFINE(M_EXT2NODE, "ext2_node", "EXT2 vnode private part");
79 static MALLOC_DEFINE(M_EXT2MNT, "ext2_mount", "EXT2 mount structure");
81 static struct vfsops ext2fs_vfsops = {
82 .vfs_fhtovp = ext2_fhtovp,
83 .vfs_mount = ext2_mount,
84 .vfs_root = ext2_root, /* root inode via vget */
85 .vfs_statfs = ext2_statfs,
86 .vfs_sync = ext2_sync,
87 .vfs_unmount = ext2_unmount,
88 .vfs_vget = ext2_vget,
91 VFS_SET(ext2fs_vfsops, ext2fs, 0);
93 static int ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev,
95 static int compute_sb_data(struct vnode * devvp,
96 struct ext2fs * es, struct m_ext2fs * fs);
98 static const char *ext2_opts[] = { "from", "export", "acls", "noexec",
99 "noatime", "union", "suiddir", "multilabel", "nosymfollow",
100 "noclusterr", "noclusterw", "force", NULL };
108 ext2_mount(struct mount *mp)
110 struct vfsoptlist *opts;
113 struct ext2mount *ump = 0;
115 struct nameidata nd, *ndp = &nd;
118 int error, flags, len;
121 opts = mp->mnt_optnew;
123 if (vfs_filteropt(opts, ext2_opts))
126 vfs_getopt(opts, "fspath", (void **)&path, NULL);
127 /* Double-check the length of path.. */
128 if (strlen(path) >= MAXMNTLEN - 1)
129 return (ENAMETOOLONG);
132 error = vfs_getopt(opts, "from", (void **)&fspec, &len);
133 if (!error && fspec[len - 1] != '\0')
137 * If updating, check whether changing from read-only to
138 * read/write; if there is no device name, that's all we do.
140 if (mp->mnt_flag & MNT_UPDATE) {
144 if (fs->e2fs_ronly == 0 &&
145 vfs_flagopt(opts, "ro", NULL, 0)) {
146 error = VFS_SYNC(mp, MNT_WAIT);
150 if (mp->mnt_flag & MNT_FORCE)
152 error = ext2_flushfiles(mp, flags, td);
153 if ( error == 0 && fs->e2fs_wasvalid && ext2_cgupdate(ump, MNT_WAIT) == 0) {
154 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
155 ext2_sbupdate(ump, MNT_WAIT);
158 vfs_flagopt(opts, "ro", &mp->mnt_flag, MNT_RDONLY);
161 g_access(ump->um_cp, 0, -1, 0);
165 if (!error && (mp->mnt_flag & MNT_RELOAD))
166 error = ext2_reload(mp, td);
169 devvp = ump->um_devvp;
170 if (fs->e2fs_ronly && !vfs_flagopt(opts, "ro", NULL, 0)) {
171 if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0))
175 * If upgrade to read-write by non-root, then verify
176 * that user has necessary permissions on the device.
178 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
179 error = VOP_ACCESS(devvp, VREAD | VWRITE,
182 error = priv_check(td, PRIV_VFS_MOUNT_PERM);
184 VOP_UNLOCK(devvp, 0);
187 VOP_UNLOCK(devvp, 0);
190 error = g_access(ump->um_cp, 0, 1, 0);
196 if ((fs->e2fs->e2fs_state & E2FS_ISCLEAN) == 0 ||
197 (fs->e2fs->e2fs_state & E2FS_ERRORS)) {
198 if (mp->mnt_flag & MNT_FORCE) {
200 "WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt);
203 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
208 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;
209 (void)ext2_cgupdate(ump, MNT_WAIT);
212 mp->mnt_flag &= ~MNT_RDONLY;
215 if (vfs_flagopt(opts, "export", NULL, 0)) {
216 /* Process export requests in vfs_mount.c. */
222 * Not an update, or updating the name: look up the name
223 * and verify that it refers to a sensible disk device.
227 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
228 if ((error = namei(ndp)) != 0)
230 NDFREE(ndp, NDF_ONLY_PNBUF);
233 if (!vn_isdisk(devvp, &error)) {
239 * If mount by non-root, then verify that user has necessary
240 * permissions on the device.
242 * XXXRW: VOP_ACCESS() enough?
245 if ((mp->mnt_flag & MNT_RDONLY) == 0)
247 error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
249 error = priv_check(td, PRIV_VFS_MOUNT_PERM);
255 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
256 error = ext2_mountfs(devvp, mp);
258 if (devvp != ump->um_devvp) {
260 return (EINVAL); /* needs translation */
272 * Note that this strncpy() is ok because of a check at the start
275 strncpy(fs->e2fs_fsmnt, path, MAXMNTLEN);
276 fs->e2fs_fsmnt[MAXMNTLEN - 1] = '\0';
277 vfs_mountedfrom(mp, fspec);
282 ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, int ronly)
285 if (es->e2fs_magic != E2FS_MAGIC) {
286 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
287 devtoname(dev), es->e2fs_magic, E2FS_MAGIC);
290 if (es->e2fs_rev > E2FS_REV0) {
291 if (es->e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP) {
293 "WARNING: mount of %s denied due to unsupported optional features\n",
298 (es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP)) {
299 printf("WARNING: R/W mount of %s denied due to "
300 "unsupported optional features\n", devtoname(dev));
308 * This computes the fields of the ext2_sb_info structure from the
309 * data in the ext2_super_block structure read in.
312 compute_sb_data(struct vnode *devvp, struct ext2fs *es,
317 int logic_sb_block = 1; /* XXX for now */
320 fs->e2fs_bsize = EXT2_MIN_BLOCK_SIZE << es->e2fs_log_bsize;
321 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
322 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
323 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
324 fs->e2fs_blocksize_bits = es->e2fs_log_bsize + 10;
325 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize;
327 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
328 fs->e2fs_bpg = es->e2fs_bpg;
329 fs->e2fs_fpg = es->e2fs_fpg;
330 fs->e2fs_ipg = es->e2fs_ipg;
331 if (es->e2fs_rev == E2FS_REV0) {
332 fs->e2fs_first_inode = E2FS_REV0_FIRST_INO;
333 fs->e2fs_isize = E2FS_REV0_INODE_SIZE ;
335 fs->e2fs_first_inode = es->e2fs_first_ino;
336 fs->e2fs_isize = es->e2fs_inode_size;
339 * Simple sanity check for superblock inode size value.
341 if (fs->e2fs_isize < E2FS_REV0_INODE_SIZE ||
342 fs->e2fs_isize > fs->e2fs_bsize ||
343 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
344 printf("EXT2-fs: invalid inode size %d\n",
349 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
350 fs->e2fs_itpg = fs->e2fs_ipg /fs->e2fs_ipb;
351 fs->e2fs_descpb = fs->e2fs_bsize / sizeof (struct ext2_gd);
352 /* s_resuid / s_resgid ? */
353 fs->e2fs_gcount = (es->e2fs_bcount - es->e2fs_first_dblock +
354 EXT2_BLOCKS_PER_GROUP(fs) - 1) / EXT2_BLOCKS_PER_GROUP(fs);
355 db_count = (fs->e2fs_gcount + EXT2_DESC_PER_BLOCK(fs) - 1) /
356 EXT2_DESC_PER_BLOCK(fs);
357 fs->e2fs_gdbcount = db_count;
358 fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize,
359 M_EXT2MNT, M_WAITOK);
360 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount * sizeof(*fs->e2fs_contigdirs),
361 M_EXT2MNT, M_WAITOK);
364 * Adjust logic_sb_block.
365 * Godmar thinks: if the blocksize is greater than 1024, then
366 * the superblock is logically part of block zero.
368 if(fs->e2fs_bsize > SBSIZE)
370 for (i = 0; i < db_count; i++) {
371 error = bread(devvp ,
372 fsbtodb(fs, logic_sb_block + i + 1 ),
373 fs->e2fs_bsize, NOCRED, &bp);
375 free(fs->e2fs_gd, M_EXT2MNT);
379 e2fs_cgload((struct ext2_gd *)bp->b_data,
381 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
386 fs->e2fs_total_dir = 0;
387 for (i=0; i < fs->e2fs_gcount; i++){
388 fs->e2fs_total_dir += fs->e2fs_gd[i].ext2bgd_ndirs;
389 fs->e2fs_contigdirs[i] = 0;
391 if (es->e2fs_rev == E2FS_REV0 ||
392 (es->e2fs_features_rocompat & EXT2F_ROCOMPAT_LARGEFILE) == 0)
393 fs->e2fs_maxfilesize = 0x7fffffff;
395 fs->e2fs_maxfilesize = 0x7fffffffffffffff;
400 * Reload all incore data for a filesystem (used after running fsck on
401 * the root filesystem and finding things to fix). The filesystem must
402 * be mounted read-only.
404 * Things to do to update the mount:
405 * 1) invalidate all cached meta-data.
406 * 2) re-read superblock from disk.
407 * 3) re-read summary information from disk.
408 * 4) invalidate all inactive vnodes.
409 * 5) invalidate all cached file data.
410 * 6) re-read inode data for all active vnodes.
411 * XXX we are missing some steps, in particular # 3, this has to be reviewed.
414 ext2_reload(struct mount *mp, struct thread *td)
416 struct vnode *vp, *mvp, *devvp;
423 if ((mp->mnt_flag & MNT_RDONLY) == 0)
426 * Step 1: invalidate all cached meta-data.
428 devvp = VFSTOEXT2(mp)->um_devvp;
429 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
430 if (vinvalbuf(devvp, 0, 0, 0) != 0)
431 panic("ext2_reload: dirty1");
432 VOP_UNLOCK(devvp, 0);
435 * Step 2: re-read superblock from disk.
436 * constants have been adjusted for ext2
438 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
440 es = (struct ext2fs *)bp->b_data;
441 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
443 return (EIO); /* XXX needs translation */
445 fs = VFSTOEXT2(mp)->um_e2fs;
446 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
448 if((error = compute_sb_data(devvp, es, fs)) != 0) {
453 if (fs->fs_sbsize < SBSIZE)
454 bp->b_flags |= B_INVAL;
460 MNT_VNODE_FOREACH(vp, mp, mvp) {
462 if (vp->v_iflag & VI_DOOMED) {
468 * Step 4: invalidate all cached file data.
470 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
471 MNT_VNODE_FOREACH_ABORT(mp, mvp);
474 if (vinvalbuf(vp, 0, 0, 0))
475 panic("ext2_reload: dirty2");
478 * Step 5: re-read inode data for all active vnodes.
481 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
482 (int)fs->e2fs_bsize, NOCRED, &bp);
486 MNT_VNODE_FOREACH_ABORT(mp, mvp);
489 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data +
490 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
501 * Common code for mount and mountroot.
504 ext2_mountfs(struct vnode *devvp, struct mount *mp)
506 struct ext2mount *ump;
510 struct cdev *dev = devvp->v_rdev;
511 struct g_consumer *cp;
516 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
517 /* XXX: use VOP_ACESS to check FS perms */
520 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
523 VOP_UNLOCK(devvp, 0);
527 /* XXX: should we check for some sectorsize or 512 instead? */
528 if (((SBSIZE % cp->provider->sectorsize) != 0) ||
529 (SBSIZE < cp->provider->sectorsize)) {
538 bo = &devvp->v_bufobj;
540 bo->bo_ops = g_vfs_bufops;
541 if (devvp->v_rdev->si_iosize_max != 0)
542 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
543 if (mp->mnt_iosize_max > MAXPHYS)
544 mp->mnt_iosize_max = MAXPHYS;
548 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
550 es = (struct ext2fs *)bp->b_data;
551 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
552 error = EINVAL; /* XXX needs translation */
555 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
556 (es->e2fs_state & E2FS_ERRORS)) {
557 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
559 "WARNING: Filesystem was not properly dismounted\n");
562 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
567 ump = malloc(sizeof *ump, M_EXT2MNT, M_WAITOK);
568 bzero((caddr_t)ump, sizeof *ump);
571 * I don't know whether this is the right strategy. Note that
572 * we dynamically allocate both an ext2_sb_info and an ext2_super_block
573 * while Linux keeps the super block in a locked buffer.
575 ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
576 M_EXT2MNT, M_WAITOK);
577 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
578 M_EXT2MNT, M_WAITOK);
579 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
580 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
581 if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
587 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
590 * If the fs is not mounted read-only, make sure the super block is
591 * always written back on a sync().
593 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
595 fs->e2fs_fmod = 1; /* mark it modified */
596 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */
599 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
600 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
601 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
603 mp->mnt_flag |= MNT_LOCAL;
607 ump->um_devvp = devvp;
608 ump->um_bo = &devvp->v_bufobj;
612 * Setting those two parameters allowed us to use
613 * ufs_bmap w/o changse!
615 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
616 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
617 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
619 ext2_sbupdate(ump, MNT_WAIT);
621 * Initialize filesystem stat information in mount struct.
624 mp->mnt_kern_flag |= MNTK_MPSAFE | MNTK_LOOKUP_SHARED |
625 MNTK_EXTENDED_SHARED;
639 mtx_destroy(EXT2_MTX(ump));
640 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
641 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
642 free(ump->um_e2fs->e2fs, M_EXT2MNT);
643 free(ump->um_e2fs, M_EXT2MNT);
644 free(ump, M_EXT2MNT);
651 * Unmount system call.
654 ext2_unmount(struct mount *mp, int mntflags)
656 struct ext2mount *ump;
658 int error, flags, ronly;
661 if (mntflags & MNT_FORCE) {
662 if (mp->mnt_flag & MNT_ROOTFS)
666 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
670 ronly = fs->e2fs_ronly;
671 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
672 if (fs->e2fs_wasvalid)
673 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
674 ext2_sbupdate(ump, MNT_WAIT);
679 g_vfs_close(ump->um_cp);
682 vrele(ump->um_devvp);
683 free(fs->e2fs_gd, M_EXT2MNT);
684 free(fs->e2fs_contigdirs, M_EXT2MNT);
685 free(fs->e2fs, M_EXT2MNT);
687 free(ump, M_EXT2MNT);
690 mp->mnt_flag &= ~MNT_LOCAL;
696 * Flush out all the files in a filesystem.
699 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
703 error = vflush(mp, 0, flags, td);
707 * Get file system statistics.
710 ext2_statfs(struct mount *mp, struct statfs *sbp)
712 struct ext2mount *ump;
714 uint32_t overhead, overhead_per_group, ngdb;
719 if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
720 panic("ext2fs_statvfs");
723 * Compute the overhead (FS structures)
726 1 /* block bitmap */ +
727 1 /* inode bitmap */ +
729 overhead = fs->e2fs->e2fs_first_dblock +
730 fs->e2fs_gcount * overhead_per_group;
731 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
732 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
733 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
738 ngroups = fs->e2fs_gcount;
740 ngdb = fs->e2fs_gdbcount;
741 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
742 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
743 ngdb += fs->e2fs->e2fs_reserved_ngdb;
744 overhead += ngroups * (1 /* superblock */ + ngdb);
746 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
747 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
748 sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead;
749 sbp->f_bfree = fs->e2fs->e2fs_fbcount;
750 sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount;
751 sbp->f_files = fs->e2fs->e2fs_icount;
752 sbp->f_ffree = fs->e2fs->e2fs_ficount;
757 * Go through the disk queues to initiate sandbagged IO;
758 * go through the inodes to write those that have been modified;
759 * initiate the writing of the super block if it has been modified.
761 * Note: we are always called with the filesystem marked `MPBUSY'.
764 ext2_sync(struct mount *mp, int waitfor)
766 struct vnode *mvp, *vp;
769 struct ext2mount *ump = VFSTOEXT2(mp);
771 int error, allerror = 0;
775 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
776 printf("fs = %s\n", fs->e2fs_fsmnt);
777 panic("ext2_sync: rofs mod");
781 * Write back each (modified) inode.
785 MNT_VNODE_FOREACH(vp, mp, mvp) {
787 if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
794 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
795 (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
796 waitfor == MNT_LAZY)) {
801 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
804 if (error == ENOENT) {
805 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
810 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
819 * Force stale file system control information to be flushed.
821 if (waitfor != MNT_LAZY) {
822 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
823 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
825 VOP_UNLOCK(ump->um_devvp, 0);
829 * Write back modified superblock.
831 if (fs->e2fs_fmod != 0) {
833 fs->e2fs->e2fs_wtime = time_second;
834 if ((error = ext2_cgupdate(ump, waitfor)) != 0)
841 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
842 * in from disk. If it is in core, wait for the lock bit to clear, then
843 * return the inode locked. Detection and handling of mount points must be
844 * done by the calling routine.
847 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
851 struct ext2mount *ump;
860 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
861 if (error || *vpp != NULL)
868 * If this malloc() is performed after the getnewvnode()
869 * it might block, leaving a vnode with a NULL v_data to be
870 * found by ext2_sync() if a sync happens to fire right then,
871 * which will cause a panic because ext2_sync() blindly
872 * dereferences vp->v_data (as well it should).
874 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
876 /* Allocate a new vnode/inode. */
877 if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
879 free(ip, M_EXT2NODE);
884 ip->i_e2fs = fs = ump->um_e2fs;
888 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
889 error = insmntque(vp, mp);
891 free(ip, M_EXT2NODE);
895 error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
896 if (error || *vpp != NULL)
899 /* Read in the disk contents for the inode, copy into the inode. */
900 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
901 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
903 * The inode does not contain anything useful, so it would
904 * be misleading to leave it on its hash chain. With mode
905 * still zero, it will be unlinked and returned to the free
913 /* convert ext2 inode to dinode */
914 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
915 ino_to_fsbo(fs, ino)), ip);
916 ip->i_block_group = ino_to_cg(fs, ino);
917 ip->i_next_alloc_block = 0;
918 ip->i_next_alloc_goal = 0;
919 ip->i_prealloc_count = 0;
920 ip->i_prealloc_block = 0;
923 * Now we want to make sure that block pointers for unused
924 * blocks are zeroed out - ext2_balloc depends on this
925 * although for regular files and directories only
927 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
928 used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
929 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
933 ext2_print_inode(ip);
938 * Initialize the vnode from the inode, check for aliases.
939 * Note that the underlying vnode may have changed.
941 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
948 * Finish inode initialization now that aliasing has been resolved.
950 ip->i_devvp = ump->um_devvp;
953 * Set up a generation number for this inode if it does not
954 * already have one. This should only happen on old filesystems.
956 if (ip->i_gen == 0) {
957 ip->i_gen = random() / 2 + 1;
958 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
959 ip->i_flag |= IN_MODIFIED;
966 * File handle to vnode
968 * Have to be really careful about stale file handles:
969 * - check that the inode number is valid
970 * - call ext2_vget() to get the locked inode
971 * - check for an unallocated inode (i_mode == 0)
972 * - check that the given client host has export rights and return
973 * those rights via. exflagsp and credanonp
976 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
984 ufhp = (struct ufid *)fhp;
985 fs = VFSTOEXT2(mp)->um_e2fs;
986 if (ufhp->ufid_ino < ROOTINO ||
987 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
990 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
996 if (ip->i_mode == 0 ||
997 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1003 vnode_create_vobject(*vpp, 0, curthread);
1008 * Write a superblock and associated information back to disk.
1011 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1013 struct m_ext2fs *fs = mp->um_e2fs;
1014 struct ext2fs *es = fs->e2fs;
1018 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1019 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1020 if (waitfor == MNT_WAIT)
1026 * The buffers for group descriptors, inode bitmaps and block bitmaps
1027 * are not busy at this point and are (hopefully) written by the
1028 * usual sync mechanism. No need to write them here.
1033 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1035 struct m_ext2fs *fs = mp->um_e2fs;
1037 int i, error = 0, allerror = 0;
1039 allerror = ext2_sbupdate(mp, waitfor);
1040 for (i = 0; i < fs->e2fs_gdbcount; i++) {
1041 bp = getblk(mp->um_devvp, fsbtodb(fs,
1042 fs->e2fs->e2fs_first_dblock +
1043 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0);
1044 e2fs_cgsave(&fs->e2fs_gd[
1045 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1046 (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1047 if (waitfor == MNT_WAIT)
1053 if (!allerror && error)
1058 * Return the root of a filesystem.
1061 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1066 error = VFS_VGET(mp, (ino_t)ROOTINO, LK_EXCLUSIVE, &nvp);